Research On Improved Multi-view Fuzzy C-Means Clustering Based On Information Entropy And Multiple Kernel Learning

At present,in fuzzy cluster field,Fuzzy C-Means(FCM)based on objective function has theoretical significance as well as practical application.Fuzzy C-Means(FCM)can be regarded as an optimization problem with membership constraints.It optimizes the objective function under the constraints,so that the sample data set can finally obtain the optimal clustering result.In view of the shortcomings of Fuzzy C-Means clustering algorithm,which is sensitive to outliers and does not perform well for non convex clustering structure,the existing IKFCM algorithm improves these two problems and enhances the robustness of the algorithm.In this paper,the convergence of IKFCM algorithm is verified by the first experiment based on the simulation data of artificial synthesis.In recent years,multi-view data are often used to fully reflect the information of data,so multi-view clustering has become an important field of statistical learning.However,in multi-view clustering,the contribution of each view to the clustering results is regarded as equally important,which is often inconsistent with the actual situation.Firstly,in order to improve this problem,this paper proposes a multi-view IKFCM clustering algorithm(WMvIKFCM algorithm),which is based on information entropy and multiple kernel learning.Especially,information entropy can automatically alter the weight of view,dramatically reduce the influence of irrelevant perspective or perspective with low contribution to clustering,and finally improve the performance of clustering,Senondly,in order to ensure the effectiveness of WMvIKFCM clustering algorithm,this paper uses Zangwill convergence theorem to analyze and prove the convergence of WMvIKFCM algorithm.The results show that:(1)when certain conditions are satisfied,the WMvIKFCM clustering algorithm produces an iterative sequence that converges or at least has a sub sequence that converges to a point in the solution set ?;(2)WMvIKFCM algorithm can find a relatively weak optimal solution in a limited number of steps.Finally,the efficiency of WMvIKFCM algorithm is verified by the second experiment based on a real dataset in the UCI database.